Financial institution leverages generative AI to automate compliance testing

Client background

The client, a large financial institution, must conduct regular risk assessments to ensure their Bank Secrecy Act (BSA) compliance programs align with regulations set forth by both the Office of Foreign Assets Control (OFAC) and the Financial Crimes Enforcement Network (FinCEN). These comprehensive assessments should quantify the risk profile of the bank's anti-money laundering (AML) programs. This process helps identify potential vulnerabilities, gaps or areas for improvement to help ensure continued compliance.

The business challenge

Data collection of unstructured and semi-structured data sources, including gathering employee census data, specific training requirements by tenure and position, and training completion certifications was required. However, manually updating and maintaining these certifications with traditional record-keeping methods proved cumbersome and lacked a consistent approach, exposing the financial institution to additional risk.

The technical team's challenge was to design a solution that could handle several aspects: frequent arrival of a large variety of unstructured and semi-structured data (such as PDFs, Excel and Word documents), processing and verifying large volumes of data for each employee, maintaining a frequent processing schedule to allow decision-makers ample time to address training gaps, and finally, ensuring user-friendliness for the non-technical staff responsible for uploading and reviewing documents.

Strategy and solution

The solution involved designing and deploying an AI Agent equipped with two key capabilities:

• Intelligent Data Extraction: This capability allows the agent to identify column headings based on a layperson's description of the specific data element. It then employs optical character recognition (OCR) to extract relevant data attributes from existing operational reports in PDF format.
• Automated Data Mapping: The AI Agent creates source-to-target column mappings, enabling the loading of previously unstructured data attributes into a structured, targeted Business Intelligence (BI) data model. Additionally, it leverages a large language model to enhance the accuracy of handwriting recognition.

To empower users and streamline analysis, the solution incorporates cloud-based Business Intelligence (BI) reports. These reports, automatically generated from the AI Agent's findings, allow staff to readily identify training gaps and employee compliance levels. The workflow is efficient: staff simply upload documents and operational reports related to employee training (certificates, transcripts, feedback forms, etc.) to the system.

Subsequently, they can leverage the user-friendly BI dashboards and reports for several key tasks, including efficiently monitoring the AI Agent's results and generating point-in-time audit reports for stakeholders. This eliminates the need for manual data analysis and simplifies the process of ensuring training compliance.

This solution led to the following key benefits:

• Reduced manual work and human errors: Streamlined document processing and verification, minimizing manual effort and potential errors.
• Enhanced responsiveness to training gaps: Provided weekly audits, allowing decision-makers to promptly address any identified training deficiencies.
• Increased compliance reporting and auditing efficiency: Improved the efficiency of both compliance reporting and audit processes.
• Regulatory compliance and risk mitigation: Ensured the client met regulatory standards, minimizing risks associated with potential penalties or reputational damage.